JP6792722B2 - Vehicle number measurement system - Google Patents

Description

The present invention relates to a technique for processing an image of a surveillance camera installed on a road and measuring the number of vehicles passing through a target section of the road.

Conventionally, various techniques related to monitoring and control of vehicles traveling on a road have been proposed. For example, Patent Document 1 discloses a technique for assisting the driving of a vehicle at a confluence such as an IC on an expressway. Further, Patent Document 2 discloses a technique for monitoring the speed of each vehicle traveling on a highway. Further, Patent Document 3 discloses a technique for centrally monitoring and managing the state of each vehicle traveling on a highway.

Japanese Unexamined Patent Publication No. 2004-102655 Japanese Unexamined Patent Publication No. 2003-346383 Japanese Unexamined Patent Publication No. 2003-337992

In facilities such as expressways, surveillance cameras are installed at regular intervals for road surveillance purposes. The measurement of the number of vehicles passing by and the speed using the image of the surveillance camera has an advantage that the equipment cost is low as compared with the method using other physical sensors.
However, the measurement of the number of passing vehicles by the image analysis of the surveillance camera has a problem that the accuracy is lowered due to the overlap of the vehicles on the image. Most of the cameras installed for surveillance purposes have their angles of view adjusted so that they can easily grasp the local traffic conditions, and they are not necessarily installed to measure the number of vehicles. For example, when monitoring a plurality of lanes with one camera, the vehicle on the back side may be hidden by the vehicle traveling in front. In this case, the vehicle traveling on the back side cannot be counted from the image of the surveillance camera.

The present invention has been made in view of the above-mentioned conventional circumstances, and is a technique capable of improving the measurement accuracy of the number of passing vehicles using the images of surveillance cameras installed along the road. The purpose is to make a proposal.

In the present invention, in order to achieve the above object, the vehicle number measurement system is configured as follows.
That is, in a vehicle number measuring system that measures the number of vehicles traveling on a road, a vehicle is detected from a plurality of photographing means installed at different positions along a predetermined section of the road and images taken by the photographing means. , The image analysis means for acquiring the speed and the feature amount of the vehicle, and the vehicle traveling in the predetermined section based on the speed and the feature amount of each vehicle in a plurality of images taken at different positions, and the predetermined It is characterized by being provided with a vehicle number calculation means for calculating the number of vehicles passing through the section.

Here, the vehicle number calculation means calculates an estimated range of the position of the vehicle based on the speed of the vehicle in the image taken at any position and the elapsed time from the time when the image is taken. It may be configured to be used.

Further, the vehicle number calculation means has a management table for managing vehicle information including the feature amount and the estimated range of the position of the vehicle being tracked, and the feature amount of the vehicle detected by the image analysis means and the vehicle feature amount. The position may be compared with the vehicle information in the management table to track the vehicle traveling in the predetermined section.

According to the present invention, it is possible to improve the measurement accuracy of the number of passing vehicles by using the image of the surveillance camera installed along the road.

It is a figure which shows the structural example of the vehicle number measuring system which concerns on one Embodiment of this invention. It is a figure which shows the arrangement example of the surveillance camera in the vehicle number measurement system of FIG. It is a figure which shows the example of the processing flow of the image analysis part in the vehicle number measurement system of FIG. It is a figure explaining the vehicle detection by the vehicle number measuring system of FIG. It is a figure which shows the example of the management table in the vehicle number measurement system of FIG. It is a figure which shows the example of the processing flow of the vehicle number estimation part in the vehicle number measurement system of FIG.

The vehicle number measuring system according to the embodiment of the present invention will be described with reference to the drawings.
FIG. 1 is a block diagram showing a configuration example of a vehicle number measurement system according to an embodiment of the present invention. In the vehicle number measurement system of this example, N units (N is an arbitrary number of 2 or more) and N associated with the surveillance cameras 1-1-1-N, respectively. It includes two video analysis units 2-1 to 2-N and a vehicle number estimation unit 3.

As shown in the arrangement example in FIG. 2, the surveillance cameras 1-1 to 1-N are set at intervals along a predetermined target section (section to be monitored) set for the road. It is assumed that there is. The spacing between adjacent surveillance cameras may be arbitrary, but shall be known.
The image analysis units 2-1 to 2-N analyze the images taken by the corresponding surveillance cameras, detect the vehicle, and extract the information of the vehicle.
The vehicle number estimation unit 3 estimates the number of vehicles that have passed through the target section based on the vehicle information detected by the video analysis units 2-1 to 2-N.

Here, the target section is a section used for measuring the number of passing vehicles, and a section without branching or merging is mainly set in the middle. As an example, a section from the entrance to the exit of a tunnel on an expressway or the like is set, and in this case, the number of passing vehicles is measured using a plurality of surveillance cameras installed at intervals in the tunnel.

The processing contents of the video analysis units 2-1 to 2-N will be described with reference to the processing flow illustrated in FIG.
First, one frame of the output video (video taken by the surveillance camera) from the corresponding surveillance camera is captured (step S101).
Next, a region in which an object exists (hereinafter referred to as an “object region”) is extracted from the captured frame (image) by a technique such as background subtraction (step S102). Specifically, for example, a background model image is generated based on past images of a plurality of frames, and an object area is extracted from the difference between the captured frame and the background model image.

Next, object tracking in the image is performed using the information of the object area extracted in step S102 and the object area extracted from the past image taken by the same surveillance camera (step S103). For example, by managing the position of the object area in the frame in association with each other, it is possible to track the moving object in the image.

Next, the feature amount is extracted from the object being tracked and managed in association with the object (step S104). As the feature amount of the object, for example, the width and height of the object, the amount of edges, the brightness of the object, the color, and the like can be considered. If it is known that the feature amount is not accurately obtained for some reason, information to that effect may be recorded. Further, the feature amount of the object may be calculated by averaging the results obtained in each frame for a weighting time.

Next, the vehicle is detected based on the feature amount of the object being tracked (step S105). For example, as shown in FIG. 4, lines 201 and 202 are defined at different positions along the road in the image area, and when an object of a certain size or larger conflicts with both lines 201 and 202, the object is moved. Detect as a vehicle. The distance between the lines 201 and 202 is determined based on, for example, the angle of the surveillance camera, the minimum size of an object assumed to be a vehicle, and the like. By performing such processing, it is possible to reduce erroneous detection of the vehicle due to noise components. Needless to say, the vehicle may be detected by another method.

Next, the speed of the vehicle detected in step S105 is calculated (step S106). The speed of the vehicle can be calculated, for example, based on the distance between the lines 201 and 202 and the time required for the vehicle to pass through this section.
The vehicle information (position, speed, feature amount, etc.) acquired by the video analysis units 2-1 to 2-N as described above is provided to the vehicle number estimation unit 3 and the vehicle has passed through the target section. It is used to estimate the number of units.

The vehicle number estimation unit 3 manages information on all vehicles detected in the target section by using a management table as illustrated in FIG.
In FIG. 5, the "label" is a unique identifier attached to the vehicle in the target section. The "speed" is the speed of the vehicle calculated by the video analysis unit, and is updated every time the vehicle is detected by the video analysis unit. The "estimated range" indicates the position of the vehicle in the target section, and is composed of "distance 1" and "distance 2" (however, distance 1 ≤ distance 2), and the vehicle number estimation unit 3 has vehicles within that range. Treat as something to do. The "feature amount" is a feature amount of the vehicle obtained by the video analysis unit. As the "feature amount", numerical values for specifying the type of vehicle such as vehicle height, vehicle width, and vehicle color are used.

In this example, the "estimated range" is updated by the following method. While the corresponding vehicle is detected by any of the video analysis units, it is updated with the true value (actual vehicle position) calculated based on the position information of the corresponding surveillance camera and the position in the frame. In this case, the same values are set for "distance 1" and "distance 2".
On the other hand, while the corresponding vehicle cannot be detected by any of the video analysis units, it is calculated using, for example, the following formula based on the "speed" and the elapsed time from the time when the video used for the calculation was taken. Update from time to time.
Distance 1 = Distance 1+ (speed x 0.95) x elapsed time Distance 2 = distance 2+ (speed x 1.05) x elapsed time This formula assumes that the traveling speed of the vehicle is not always constant, and the speed is 5%. It is set to handle fluctuations. Needless to say, the coefficient of speed fluctuation is not limited to 5%, and for example, a coefficient corresponding to a measured value of past speed fluctuation on the target road can be used.
If any of the video analysis units can detect the corresponding vehicle again, the update is returned to the true value.

For example, when both "distance 1" and "distance 2" of the "estimated range" are "500", it means that the vehicle is detected at a position 500 m from the start point of the target section. Further, for example, when the "distance 1" and "distance 2" of the "estimated range" are "275" and "325", the positions within the range of 275 m to 325 m from the start point of the target section (that is, 300 m from the start point). It means that it is estimated that the vehicle exists at a position within a range of 50 m centered on the point.
By managing the vehicle using such a management table, not only can the vehicle be tracked while the vehicle is being photographed by one of the surveillance cameras, but the vehicle can be tracked while the vehicle cannot be photographed by any of the surveillance cameras. can do.

When the "distance 1" becomes equal to or greater than a predetermined value and it is presumed that the vehicle has left the target section, the vehicle number estimation unit 3 removes the vehicle information of the vehicle from the management table of FIG. 5, and the number of vehicles passing through the target section. The number of vehicles is measured by counting up. The vehicle information removed from the management table may be left in a log or the like. The determination of whether or not the vehicle has left the target section is not limited to the above method, and for example, it may be determined whether or not the median value of "distance 1" and "distance 2" is equal to or greater than a predetermined value. Good.

The process of tracking the vehicle number estimation unit 3 while associating the vehicles with the plurality of surveillance cameras 1-1 to 1-N will be described with reference to the process flow illustrated in FIG.
When a vehicle is detected by any of the video analysis units 2-1 to 2-N associated with the surveillance cameras 1-1 to 1-N (step S201), the vehicle number estimation unit 3 is the image analysis unit. Vehicle information (position, speed, feature amount, etc.) is acquired from and the following processing is performed.

First, vehicle information whose detected vehicle position is included in the "estimated range" is extracted from the management table as vehicle information of the candidate vehicle (step S202).
Next, the feature amount of the detected vehicle is compared with the feature amount of each of the extracted candidate vehicles, and it is determined whether or not the detected vehicle is a new vehicle (step S203). Specifically, for example, the similarity with the feature amount of the detected vehicle is calculated for each candidate vehicle, and if there is no candidate vehicle whose similarity is equal to or higher than the threshold value, it is determined as a new vehicle, and the similarity is the threshold value. If there is a candidate vehicle as described above, it is determined that the vehicle is not a new vehicle.

When it is determined that the vehicle is a new vehicle in step S203, a new label is given to the detected vehicle, and vehicle information is added to the management table (step S204).
On the other hand, when it is determined in step S203 that the vehicle is not a new vehicle, the detected vehicle is assumed to be the same vehicle as one vehicle selected from the candidate vehicles narrowed down based on the similarity of the feature amounts. , Update the vehicle information of the selected candidate vehicle (step S205). The vehicle selected from the candidate vehicles may be a vehicle having the feature amount closest to the feature amount of the detected vehicle, or the relationship between the position of the detected vehicle and the position (estimated range) of the candidate vehicle (for example, estimation). The vehicle may be determined from the proximity to the center of the range). Further, the vehicle may be selected in consideration of both the feature amount and the position of the vehicle.

As described above, the vehicle number measuring system of this example includes a plurality of surveillance cameras 1-1 to 1-N (an example of photographing means according to the present invention) installed at different positions along a predetermined section of the road. Video analysis unit 2-1 to 2-N (an example of the image analysis means according to the present invention) that detects a vehicle from the video captured by the surveillance cameras 1-1 to 1-N and acquires the speed and feature amount of the vehicle. ) And a vehicle number estimation unit that tracks vehicles traveling in a predetermined section based on the speed and feature amount of each vehicle in a plurality of images shot at different positions, and calculates the number of vehicles that have passed within the predetermined section. 3 (an example of the vehicle number calculation means according to the present invention) is provided.

In this way, the vehicle number measurement system of this example integrates the results of analyzing the images of a plurality of surveillance cameras arranged at different positions along the target section of the road, and obtains a vehicle traveling in the target section. Count the number of vehicles while tracking. That is, the number of vehicles that have passed through the target section covered by a plurality of surveillance cameras is measured instead of the imaging range of the surveillance camera alone. With such a configuration, if a vehicle is photographed by any of the surveillance cameras in the target section, the vehicle is counted. Therefore, a vehicle that cannot be photographed by one surveillance camera because it overlaps with another vehicle can be photographed by another surveillance camera, and the difficulty of counting due to the overlap of vehicles can be reduced. It is possible to measure the number of units with high accuracy.

Further, in the vehicle number measuring system of this example, the vehicle number estimation unit 3 determines the vehicle speed based on the speed of the vehicle in the image captured at any position and the elapsed time from the time when the image is captured. It is configured to calculate the estimated range of the vehicle position.
Therefore, even if there is a period during which the surveillance camera cannot take a picture of the vehicle, the tracking of the vehicle can be continued. Therefore, it is possible to prevent the same vehicle from being counted twice, and the measurement accuracy of the number of vehicles is improved.

Further, in the vehicle number measurement system of this example, the vehicle number estimation unit 3 has a management table that manages vehicle information including the feature amount and position estimation range of the vehicle being tracked, and the video analysis unit 2 The feature amount and position of the vehicle detected by -1 to 2-N are compared with the vehicle information in the management table, and the vehicle traveling in the predetermined section is tracked.
Therefore, it is possible to accurately track the vehicle based on the images taken by different surveillance cameras, and the measurement accuracy of the number of vehicles is improved.

The vehicle for which the number of vehicles is to be measured is not limited to the automobile, and may include a two-wheeled vehicle. When measuring the number of two-wheeled vehicles, it is sufficient to use attributes peculiar to the two-wheeled vehicle such as the width of the vehicle, the overall silhouette, and the number of tires as the feature quantities of the vehicle so that the two-wheeled vehicles can be individually tracked. Further, the number of passing vehicles and the number of passing motorcycles may be measured separately.

Here, the configurations of the system, the device, and the like according to the present invention are not necessarily limited to those shown above, and various configurations may be used. For example, in the above description, the video analysis unit is provided for each surveillance camera, but one video analysis unit may be configured to process images from a plurality of surveillance cameras.

Further, the present invention can be provided, for example, as a method or method for executing the process according to the present invention, a program for realizing such a method or method, a storage medium for storing the program, or the like.

The present invention can be used in a vehicle number measuring system that measures the number of vehicles traveling on a road.

1-1-1-N: Surveillance camera, 2-1-2-N: Video analysis unit, 3: Vehicle number estimation unit

Claims (3)

In a vehicle number measurement system that measures the number of vehicles traveling on the road
Multiple shooting means installed at different positions along a predetermined section of the road,
An image analysis means that detects a vehicle from an image taken by the photographing means and acquires the speed and feature amount of the vehicle.
A vehicle number calculation means that tracks vehicles traveling in the predetermined section based on the speed and feature amount of each vehicle in a plurality of images shot at different positions, and calculates the number of vehicles passing through the predetermined section. With ,
The vehicle number calculation means has a management table that manages vehicle information including an estimated range of the feature amount and position of the vehicle being tracked, and obtains the feature amount and position of the vehicle detected by the image analysis means. A vehicle number measuring system characterized in that a vehicle traveling in the predetermined section is tracked as compared with the vehicle information in the management table . In the vehicle number measuring system according to claim 1,
The vehicle number calculation means calculates an estimated range of the position of the vehicle based on the speed of the vehicle in the image taken at any position and the elapsed time from the time when the image was taken. A featured vehicle number measurement system. In the vehicle number measuring system according to claim 1 or 2 .
The vehicle number calculation means extracts vehicle information including the detected vehicle position in the estimation range as vehicle information of the candidate vehicle from the management table, and determines the degree of similarity with the feature amount of the vehicle detected for each candidate vehicle. It is characterized in that it is calculated and determined as a new vehicle when there is no candidate vehicle having the similarity equal to or higher than the threshold value, and it is determined not to be a new vehicle when there is a candidate vehicle having the similarity degree equal to or higher than the threshold value. Vehicle number measurement system.

Images